Ecological Studies Flashcards
what is an ecological study
observational study where data is analysed at population or group level rather than individual
area-level studies to measure prevalence and incidence of disease - particularly when it is rare
what are the advantages of ecological studies
inexpensive
easy to do as use routinely collected data
population context of individual characteristics is a stronger determinant of disease at population level than individual level risk factors
what are the disadvantages of ecological studies
prone to bias and confounding
care must be taken when extrapolating to individuals within the measurement areas
or to higher population level
other designs are generally considered more reliable in respect to inference and causation
why would someone do an ecological study
to monitor population health
informing and develop public health strategies
for large scale comparisons e.g between countries
to study the relationship between population-level exposure to risk factors and disease
or to study contextual effect of risk factors on population
if individual measurements are not available
if the disease is rare
whats an example of when individual measurements may not be available and one has to do an ecological study?
confidentiality precludes the release of data on individuals
so its anonymised by aggregation of data to small area level
what are the types of measurement
health outcomes (incidence/prevalence/rate)
aggregate measures (mean/%/area-level deprivation indices)
environmental measures
global measures (e.g number of GPs, or population density)
what is a geographical ecological study
compares one geography to another and assesses health of each
exposure may be measured and included in analysis
also possible confounders
e.g demographic, socioeconomic
what is a longitudinal ecological study
population monitored to assess changes in disease over time
confounding factors may be included in analysis
what is a migration ecological study
migrant population data collected and analysed
unit of interest is population type
not time or place
what is an example of ecological fallacy
falsely assumes that every individuals IQ is high, , since the class average is high
e.g if students in class got 50, 90, 130 and 150 would average 105
what is ecological fallacy
type of confounding specific to ecological studies - occurs when relationships found for groups are assumed also to be true for individual
- may be unknown confounding factors that display same pattern of spatial heterogeneity
why does ecological fallacy happen
the data is aggregated so does not provide exact information on those particular individual
how do we overcome ecological confounding?
regression modelling
poisson regression
negative binomial regression
multilevel modelling
what are area-level exposures
environmental exposures determined by area
e.g in air or drinking water
individual exposure to these things associated with area of residence work, study - reflects demographic and lifestyle factors - age, sex , SES
what is the ideal analysis and problem with it
based on individual-level data but not feasible or very expensive to obtain
what can be done instead of individual-level data
need to model exposures
but proxy measures based on models may not capture exposure of individual
what can modelling exposures lead to
exposure misclassification
what does exposure misclassification often lead to
it is non-differential to disease status so leads to
regression dilution bias - means the bias of effect size estimates towards the null
what is the excessive relative risk of environmental exposures expected to be
small
what could non-differential exposure mis-classification leading to a bias toward the null lead to
false-negative results
false reassurance about the health effect of exposure
what is the implication to public health of non-differential exposure misclassification
even small excess relative risks applied to large numbers of people could result in large excess numbers of disease
consequential for rare diseases - e.g childhood leukemia
outline the bone cancer example of ecological study
aim to test prediction of spatial variations that may arise as result of environmental mechanisms and SES deprivation
based on the density of population of area and SES
patients diagnosed 1980-2005, data from registry
what were the 4 census measures in the bone cancer study?
unemployment
households with no car
non-home ownership
household overcrowding
what were the findings of the bone cancer study
females - lower incidence of osteosarcoma associated with higher deprivation
lower incidence of Ewing sarcoma associated with living in more densely populated area and higher non-car ownership
what can be concluded for females and osteosarcoma in the bone cancer study?
geographical heterogeneity of incidence is modulated by differences in environmental exposure occurring in less and more socio-economically deprived areas of residence
what can be concluded for Ewing sarcoma in the bone cancer study?
geographical heterogenity of incidence is modulated by differences in environmental exposures occurring in less and more densely populated areas
also modulated by differences between environmental exposure occurring in less and more socioeconomically deprived areas
outline the fluoride study
it is in drinking water - possibility of bone cancer?
examine if increased risk of bone cancer is linked to residence in areas with higher fluoride in water
hypothesis - modulated for osteosarcoma but NOT Ewing sarcoma
statsitical analysis - negative binomial regression - with adjustments for previously identified deprivation
what were the findings of the fluoride study
no evidence for the association between fluoride in water and osteosarcoma
no evidence for Ewing sarcoma either
why may there have been no association found in the fluoride study?
because of attenuation due to exposure measurement error - arising through imprecision of allocating fluoride levels
the study was ecological and so may not reflect individual exposures
what is the type 1 diabetes study
incidence relationship between childhood type 1 and levels of nitrate in drinking water
yorkshire
statsitical model was poisson instead of negative binomial
what did the type 1 diabetes study find
diabetes incidence was positively associated with raised mean nitrate levels
significant negative trend found between standardised incidence ratio and proportion of non-caucasians in population, and population density
what did the type 1 diabetes study conclude
findings were not explained by ethnic composition of pop, pop density or SES
nitrate in drinking water may be a precursor of chemicals toxic to the pancreas
